Currently the server searches for denormalized hash ids for each client after topology change / new client joining.

This is pretty expensive operation as it results in calling the hash algorithm several hundred thousand times, and this for each client.
The result of this operation should be cached.